GEO

Does llms.txt Actually Work? What the Data Says (2026)

Some links here are affiliate links: if you buy through them I may earn a commission, at no extra cost to you. I only recommend tools I'd point a colleague to, and rankings are never paid for.

On this page
  1. What llms.txt actually is
  2. What the data says
  3. Why it doesn’t (yet) work
  4. The honest verdict
  5. Measure, don’t guess

Short answer: based on the best data available, llms.txt does not measurably improve your AI citations. SE Ranking analyzed roughly 300,000 domains and found no statistically significant correlation between having an llms.txt file and how often AI engines cite a site — in fact, removing llms.txt from their citation-prediction model improved its accuracy, meaning the file was noise, not signal. Only one of the 50 most-cited domains even had one. llms.txt is free and harmless to add, and it may help you later if AI agents adopt it, but it is not the citation lever many people claim. Here’s what it is, what the data actually says, and where to put your effort instead — from an SEO of eleven years running his own site through this category in the open.

What llms.txt actually is

llms.txt is a proposed standard: a Markdown file at the root of your domain (like robots.txt, but for language models) that lists your most important pages and content in a clean, LLM-friendly format. The idea is reasonable — give AI models a curated map of your site so they can find and understand your best material without wading through nav, ads and scripts.

The key word is proposed. It’s a community proposal, not something the major AI companies have committed to using. That distinction is the whole story.

What the data says

The most rigorous look so far is SE Ranking’s analysis of ~300,000 domains (their write-up, covered by Search Engine Journal). The findings are blunt:

  • No measurable link between llms.txt adoption and AI citation frequency.
  • Removing llms.txt as a variable made their predictive model more accurate — a strong sign the file adds no signal.
  • Adoption is low: about 10.13% of domains had an llms.txt file.
  • Among the 50 most AI-cited domains, only one had llms.txt at all.

A separate study by Trakkr scanned 37,894 domains and reached the same place — zero citation advantage. SE Ranking’s own conclusion is measured: llms.txt “doesn’t seem to directly impact AI citation frequency. At least not yet.”

Why it doesn’t (yet) work

The reason is simple: no major AI engine has confirmed it reads llms.txt. OpenAI, Google and Anthropic crawl and cite the open web the way they already do — through the same content and link signals that drive their answers. A file they haven’t agreed to consume can’t influence citations. Until an engine says “we use this,” llms.txt is a bet on a future that may or may not arrive.

Pros

  • Free and quick to add — near-zero downside
  • A tidy, curated map of your site that may help future AI agents
  • Forces you to identify your most important pages (a useful exercise)

Cons

  • No measurable effect on AI citations across 300k domains
  • No major AI engine has confirmed it uses the file
  • Easy to over-invest in as a "hack" while ignoring the real levers
  • Adoption is low and concentrated away from the most-cited sites

The honest verdict

Add llms.txt if you want — it costs nothing, it won’t hurt, and it’s a sensible hedge if AI agents start honoring it. (This site has one, and I built a free llms.txt generator if you want to make a valid one in a minute.) But don’t treat it as a growth tactic, and definitely don’t pay anyone for an “llms.txt strategy.” The effort that actually moves your AI visibility goes into the things the models do reward: genuinely quotable content, clear entity signals, and third-party mentions on sources the models already trust. I wrote that playbook here: how to get cited by ChatGPT — and you can score any page’s citability for free to see where it stands.

Measure, don’t guess

The deeper lesson from the llms.txt hype cycle is to stop trusting tactics and start measuring outcomes. Instead of assuming a file helped, track whether AI engines actually cite you — before and after any change — with a monitoring tool. That’s the only way to know what’s working.

Track your real AI citations with RankScale

For the full set of tools that measure this, with pricing and who each fits, see my best GEO tools comparison. New to the shift? Start with GEO vs SEO: what actually changes when AI answers the query. Not sure which tool suits you? The 60-second GEO tool finder narrows it to one.

Frequently asked questions

Does llms.txt help you get cited by AI?

Based on the largest study available — SE Ranking's analysis of about 300,000 domains — no. There is no measurable correlation between having an llms.txt file and AI citation frequency, and removing it from their prediction model actually improved accuracy. It is free and harmless to add, but it is not the citation lever many claim.

What is llms.txt?

It is a proposed standard: a Markdown file at your domain root that lists your most important pages in a clean, LLM-friendly format — like robots.txt, but aimed at language models. It is a community proposal, not something the major AI companies have committed to using.

Do ChatGPT, Google or Perplexity use llms.txt?

None of the major AI engines has confirmed that it reads llms.txt. They crawl and cite the open web through the same content and link signals that drive their answers. That is why the file has no measurable effect on citations today.

Should I still add llms.txt?

You can — it costs almost nothing, it will not hurt, and it is a reasonable hedge if AI agents start honoring it in future. Just do not expect it to earn you citations, and do not pay for an "llms.txt strategy." Put real effort into quotable content, entity signals and third-party mentions instead.

What actually gets you cited by AI instead?

The levers the models reward: content that is genuinely quotable and well-structured, clear entity signals so models understand who you are, and mentions on third-party sources the models already trust. Then measure it — track whether AI engines cite you before and after changes with a monitoring tool.

Published Last updated

← All posts

navigate openesc close